HSV Genital Skin Biopsy scRNAseq Explorer

By: Joe Hou

Overview

This Shiny application provides an interactive interface for exploring 10X scRNAseq data derived from HSV+ skin biopsy samples. https://hsvdashboard.joehou.net/.
As the manuscript is currently under review for publication, the raw and processed data remain undisclosed.

- About the cohort:

This study involved healthy adults with confirmed herpes simplex virus type 2 (HSV-2) seropositivity, verified via Western blot analysis. All participants tested negative for HIV. In this longitudinal HSV study, we enrolled 17 participants and collected genital skin biopsies at three distinct time points to capture HSV shedding and healing phases.

- About the time points:

Samples were collected at three key stages: the first, labeled “Prior,” was collected before any visible lesions appeared. When lesions became visible, samples were labeled as “Lesion.” Finally, samples taken 8 weeks after lesion appearance are referred to as “Post.”

- About the dataset:

This release includes datasets with refined mapping for T cells, Myeloid cells and Visium data. Users can freely select from the available datasets for exploration. We recommend using the “CellType_Level3” annotation for visualization, as it provides the most detailed and granular clustering. While “CellType_Level1” offers a broader categorization, “CellType_Level3” represents the final, detailed clustering, with cell types identified at a higher level of specificity. Full definitions of each cell cluster and cell type will be provided in the manuscript upon publication.

Features and Functionalities

Two main tabs are provided for scRNAseq and spatial Visium data.

For scRNAseq data:

Two separate tabs are provided for data exploration, with a focus on:

  1. Cell Cluster Composition:
  • Goal: Analyze the composition of cell clusters across time points and individuals, with some clusters showing specificity to time points and variability across participants.
  1. Gene Expression Patterns:
  • Goal: Investigate gene expression patterns across different cell clusters, including gene distribution and expression variability across time points.

For Visium data:

  1. Gene Expression Patterns:
  • Goal: Investigate gene expression distribution and intensity across tissue.
  1. Cell Type De-convolution:
  • Goal: To further understand cell type composition per spatial location.

Interactive Exploration on Cell Type (Cluster Discovery Tab)

  • Cell Type and Status Identification: Users can view all available cell types and subjects, with the flexibility to select specific ones for detailed exploration.
  • Dynamic UMAP Visualization: Initial displays show holistic UMAP results, which can be refined by time point. Selecting different subjects and cell types dynamically updates the UMAP to show specific clusters.
  • Cell Type Statistics: Displays percentages and counts of cell types for selected or all subjects.
  • Cell Type by Subject: Visualizes the percentage of each cell type within a given sample using stacked bar plots to understand the heterogeneity of cell type acroos samples

Interactive Exploration on Gene Expression (Gene Discovery Tab)

  • Feature Exploration: Allows for the selection of cell types, subjects, and specific genes of interest.
  • At Single Cell Level:
  • Feature Gene Highlight: The expression of selected gene projected to UMAP overlay with cell type annotation, vividly display gene expression pattern across cell types.
  • Heatmap: Shows selected gene expression intensity per cell, grouped by cell types, statuses, and subject, aiding in providing gene details per sample.
  • At Cell Type Level:
  • Violin Plots: The selected gene expression was summarized into cell type level and shows average expression levels across cell types and statuses, aiding in understanding temporal dynamics.
  • Feature Percentage Plots: Summarizes gene expression at the sample/subject level, indicating the proportion of cells expressing a specific gene within a cell type.
  • Dot Plots: Dot size indicates the percentage of cells expressing each gene, while color intensity represents gene expression levels.

Tech Note

  • HDF5 Array File System: Utilized to prevent RAM overload by saving massive data in HDF5 files and processing them directly from disk. This approach enhances memory efficiency when handling large datasets.
  • Docker Image: The application is containerized using Docker to ensure consistency across different platforms. This allows for a reliable and reproducible environment for all users. You can run command line below to get this container runs locally: finch run -it –rm -p 7777:3838 jhoufred/hsv-dashboard-image:2.0 or this: docker run -d -p 7777:3838 jhoufred/hsv-dashboard-image:2.0
  • AWS EC2 Backend: The application is deployed on AWS EC2, providing a scalable and robust backend infrastructure for performance and reliability.

Version2 Update

  • NEW AWS Architecture: Elastic Container Registry (ECR), Elastic Container Service (ECS), Application Load Balancer (ALB), Auto Scaling, and Route 53